the relative returns to credit- and non-credit-bearing credentials · 2020. 10. 5. · arellanes...

M A K I N G RES EA RC H R EL EVA N T

The Relative Returns to Credit- and

Non-Credit-Bearing Credentials October 2020

Candace Hester │ Sami Kitmitto

The Relative Returns to Credit- and

Non-Credit-Bearing Credentials OCTOBER 2020

Candace Hester | Sami Kitmitto

1400 Crystal Drive, 10th Floor Arlington, VA 22202-3239 202.403.5000

www.air.org

Notice of Trademark: “American Institutes for Research” and “AIR” are registered trademarks. All other brand, product, or company names are trademarks or registered trademarks of their respective owners.

Copyright © 2020 American Institutes for Research®. All rights reserved. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, website display, or other electronic or mechanical methods, without the prior written permission of the American Institutes for Research. For permission requests, please use the Contact Us form on www.air.org.

http://www.air.org/

We are extremely grateful for the financial support of the Education Credit Management

Corporation Foundation to produce this research. In addition, we would like to thank Melissa

Arellanes and Semret Yibass for ongoing and thoughtful research support, along with Heinrich

Hock, Scott Davis, Katherine Hughes, Alexandria Radford, and Stephen Plank for thoughtful

comments and feedback.

The Relative Return of Credit- and Non-Credit-Bearing Educational Attainment

Contents Page

Abstract ........................................................................................................................................... iii

Introduction .................................................................................................................................... 1

Background ..................................................................................................................................... 3

Why Focus on Middle Skills? ....................................................................................................... 4

The Labor Market Returns to Middle-Skill Training.................................................................... 5

Methods .......................................................................................................................................... 7

Data ............................................................................................................................................... 10

Descriptive Changes in the Outcomes of Interest: Employment and Earnings ........................ 12

Employment .............................................................................................................................. 12

Earnings ..................................................................................................................................... 14

Impact Findings ............................................................................................................................. 16

Hours and Earnings ................................................................................................................... 17

Subgroup Analyses .................................................................................................................... 19

Exploratory Analyses ................................................................................................................. 21

Conclusion ..................................................................................................................................... 22

References .................................................................................................................................... 24

Appendix A. Description of the NLSY97 Data ............................................................................... 28

Overview of the Data Source .................................................................................................... 28

Sample Characteristics .............................................................................................................. 28

Educational Attainment ............................................................................................................ 29

Employment and Earnings ........................................................................................................ 31

Weights ..................................................................................................................................... 32

Appendix B. Findings on Wages and Hours .................................................................................. 33

Appendix C. Exploratory Subgroup Analysis Results .................................................................... 35


Figures Page

Figure 1. The Flow of Schooling and Training Questions in the NLSY97 Used to Identify Educational Attainment Type and Date of Completion ................................................................ 11

Figure 2. Percentage Employed by Age and Highest Educational Attainment ............................. 13

Figure 3. Percentage Employed by Years Before and After Educational Attainment .................. 14

Figure 4. Earnings by Age and Highest Educational Attainment .................................................. 15

Figure 5. Earnings by Year Before and After Educational Attainment ......................................... 16

Tables

Table 1. NLSY97 Sample Characteristics ....................................................................................... 12

Table 2. Regression Estimates of Employment Likelihood and Natural Log Earnings by Educational Attainment ................................................................................................................ 17

Table 3. Regression Estimates of Yearly Hours Worked and Natural Log Hourly Wage by Educational Attainment ................................................................................................................ 18

Table 4. Regression Estimates of Employment Likelihood Among Vulnerable Subgroups .......... 19

Table 5. Regression Estimates of Natural Log Yearly Earnings Among Vulnerable Subgroups ..................................................................................................................................... 20

Table 6. Regression Estimates of First Attainment in Credit-Bearing Programs Compared With Non-Credit-Bearing Programs on the Propensity to Attain a Bachelor’s Degree by Age 30 ........................................................................................................................................... 22

Table B1. Regression Estimates of Yearly Hours Worked Among Vulnerable Subgroups ........... 33

Table B2. Regression Estimates of Natural Log Hourly Wage Among Vulnerable Subgroups ..................................................................................................................................... 34

Table C1. Regression Estimates of First Attainment in Credit-Bearing Programs Compared With Non-Credit-Bearing Programs on the Propensity to Attain a Bachelor’s Degree by Age 30 .......................................................................................................................... 35


Abstract

Prior research indicates that future workers will not possess enough training to contribute to a

robust U.S. labor market, especially in “middle-skill” roles that require some postsecondary

education but not a 4-year degree. At the same time, the last 30 years have borne witness to an

increased likelihood that middle-skill individuals pursue nontraditional, non-credit-bearing, skill-

based training (such as for computer operations), with little accompanying evidence on the

utility of these less portable options. To address this research void, we analyzed 20 years of the

National Longitudinal Survey of Youth 1997, using an individual fixed-effects regression strategy

to estimate the returns to non-credit-bearing credential and licensure pathways compared with

credit-bearing credential and associate degree programs. This strategy provided estimates on

the returns to training that are unrelated to individual-level, time-invariant factors, such as

differences in ability and motivation which remain constant over time. Our findings show that

credit-bearing credentials yield an approximately equal likelihood of employment as non-credit-

bearing credentials, but significantly improved earnings of about 17%, which typically equates

to about $5,500 per year.


AMERICAN INSTITUTES FOR RESEARCH® | AIR.ORG 1

Introduction

In a labor market defined by rapid technological progress, changing demographics, and fierce

global competition, jobs will increasingly require postsecondary education or training. Yet

studies suggest that future workers will not possess the training to fill these jobs (Carnevale,

Smith, & Strohl, 2013). Furthermore, training gaps are greatest among middle-skill jobs that

require some postsecondary education but not a 4-year degree (National Skills Coalition, 2017).

At the same time, the marketplace for middle-skill training is undergoing significant changes. For

example, the number of individuals completing less-than-2-year certificate programs at for-profit

colleges increased by more than 500% between 1990 and 2011, whereas the comparable statistic

at public institutions was less than 200%.1 This rapidly changing landscape and the growing

demand for middle-skill employees has left policymakers and students with limited information

on the best way to prepare workers (Deming, Goldin, & Katz, 2012, 2013; Mettler, 2014).

1 These tabulations are based on our calculations using the Integrated Postsecondary Education Data System downloaded from https://nces.ed.gov/ipeds/. Specifically, in the 1989–90 school year, 89,863 and 21,775 individuals completed less-than-2-year certificates at public and for-profit institutions, respectively. In 2010–11, the comparable statistics were 175,053 and 115,532, respectively. It also is worth noting for this particular example that for-profit college enrollment thereafter declined, fairly markedly, to about 43,212 completions by 2017–18.

https://nces.ed.gov/ipeds/



In this report, we present evidence on the relative labor market returns to more transferable

middle-skill postsecondary training, including credit-bearing certificate and associate degree

programs (hereafter, credit-bearing programs) compared with terminal but targeted, non-

credit-bearing certificate and licensure programs (hereafter, non-credit-bearing programs). In

some cases, the credit- and non-credit-bearing programs are in similar fields (e.g., nurse’s aide,

computer operations programs). However, non-credit-bearing programs do not build toward an

associate or bachelor’s degree in the same ways that credit-bearing programs do. The appeal of

many non-credit-bearing programs therefore is that they are shorter and more targeted, even

sometimes less expensive and provided directly through an employer. Alternatively, the appeal

of credit-bearing programs derives from their more portable, flexible nature: Students can earn

credits that will follow them from one program to the next, from one employer to the next,

across time.

To present evidence on the relative returns that credit-bearing and non-credit-bearing

programs yield, we used the National Longitudinal Survey of Youth 1997 (NLSY97), and followed

individuals from the time they were 18 years old (which occurred anywhere between 1996 and

2000) through the year in which they were 30 (which occurred anywhere between 2008 and

2014). We used NLSY97 records on respondents’ education and ongoing labor market

outcomes.

Our identification strategy leveraged individual fixed-effects regression models to examine

within-person changes in the likelihood of employment and earnings associated with credit-

and non-credit-bearing program completion. The innovation of the individual fixed-effects

strategy is that it controls for any differences between individuals that are fixed across time

(e.g., ability) and estimates the labor market returns associated with program completion that

accrue to an individual, relative only to prior labor market participation. To control for

differences in an individual’s likelihood to select into a program pathway based on qualities that

may change during one’s life course (e.g., labor market conditions, marital status, child rearing),

we also exploited time-varying controls on the county of residence and family makeup. Our

models also include a control for workforce participation that is coincident with school

enrollment to absorb variation associated with systematic differences in the propensity to work

during credit- and non-credit-bearing training.

Our findings show that credit-bearing programs did not yield a significant increase in the

likelihood of employment relative to non-credit-bearing programs. On the other hand, findings

did indicate that credit-bearing programs increased yearly earnings by about 17% relative to the

$33,000 average earnings of respondents who completed non-credit-bearing programs and

who were employed, representing a change of about $5,500.



We tested the consistency of these findings across subgroups that prior literature indicates are

vulnerable to exploitative training programs (e.g., from private for-profit training institutions).

These groups include individuals from low-wealth families, wherein wealth is in the lowest

tercile of the wealth distribution when the respondent was age 18; non-White individuals; and

individuals whose parents had not attained a 4-year degree by the first interview wave in

1996.2 Among these subgroups, the earnings boost to credit-bearing program completion is not

statistically significant for low-wealth and non-White respondents.

We also present preliminary

evidence that the positive returns to

credit-bearing programs are

associated with the improved

flexibility of the credential.

In the sections that follow, we

discuss the background literature,

describe the individual fixed-effects

methodology, describe our data,

present findings, and conclude with

policy implications.

Background

Technological advances are changing the nature of work in the United States. Research

indicates that these advances have, in turn, increased the demand for workers with skills

beyond a high school education, especially middle-skill training, requiring some postsecondary

education but not a 4-year degree (Marr, 2019; National Skills Coalition, 2017). Despite this

evidence showing that not everyone in the economy needs to or should attain a 4-year degree,

particularly given the high and rising costs of education (Fuller & Raman, 2017), the research on

the returns of higher education disproportionately focuses on 4-year degrees. In addition, the

research on the returns to attaining less postsecondary education, while long-standing,

generally focuses on a somewhat narrow conceptualization of middle-skill training associated

with associate degrees from higher education institutions. In addition, this prior work presents

wide variation in the results: studies found increases in annual earnings for attaining associate

degrees ranging between 15% and 27% (Abel & Deitz, 2014; Kane & Rouse, 1995). Considering

2 For an example of research showing these subgroups are vulnerable, see Le, Yang, & Simko (2017).



research on training that occurs outside higher education institutions, the studies become even

less numerous and the estimates more varied.

To better inform decisions about and investments in the pursuit of middle skills, this study

investigated one potentially important distinction among middle-skill training programs: credit-

bearing programs in which participation generates credits that can be used toward a 4-year

degree compared with terminal, non-credit-bearing programs. The credit-bearing category

includes certificate and associate degree programs. These contrast with non-credit-bearing

programs that also can lead to certificates, industry-recognized certifications, or licenses but do

not award credits for participation or completion.

We anticipated at the outset that credit-bearing programs lead to better labor market

outcomes for two reasons. First, credit-bearing programs have a requisite association with an

accredited institution and, perhaps as a result, generally span more years.3 Second, credit-

bearing programs provide students with the opportunity to transfer and stack their credentials

because of the linking role of credits across institutions and time. Therefore, credit-bearing

programs, unlike non-credit-bearing programs, provide a gateway for attaining higher degrees.

In the subsections that follow, we discuss in greater detail the sizable and growing need for

middle-skill training plus prior estimates on the return to education for middle-skill work.

Why Focus on Middle Skills?

For policymakers, a focus on middle-skill jobs is important for two reasons. First, middle-skill

jobs provide an increasingly important pathway to the middle class. As the number of well-

paying jobs available for those with a high school diploma or less shrinks, the number of jobs

for those with more education increases (Carnevale, Strohl, Ridley, & Gulish, 2018). Second, the

largest gap between the labor market demand and the labor force exists among the middle

skilled: Estimates from 2015 showed that only 16% of jobs are low-skill jobs, requiring no

postsecondary training, and only about one third of jobs (31%) are high-skill jobs, requiring a 4-

year degree. Thus, more than half of the jobs (53%) are middle-skill jobs. At the same time, only

43% of the workforce was middle skilled (National Skills Coalition, 2017). In addition, a survey of

human resource executives in 2014 found that more than two thirds of respondents felt that

“their inability to attract and retain middle-skills talent frequently affected their performance”

(Burrows, Young, Restuccia, Fuller, & Raman, 2014, p. 6).

Although the exact definition of middle-skill jobs is still a subject of debate (Handel, 2017),

these jobs typically require a credential acquired through a certificate program or an associate

3 Non-credit-bearing programs also may be associated with accredited institutions. However, the association with accredited institutions is not a requirement.



degree. These jobs are in a broad variety of fields, from manufacturing to health care to

business, and the long-term outlook for these jobs can vary significantly. A 2018 study of

middle-skill job-seeker résumés examined variance in earnings and career trajectories and

found that these jobs fall into three general categories:

Lifetime—a middle-skill job that is a career in itself, allowing for advancement within the same

role and long-term job stability (most common in the health-care field [e.g., nurses and

dental hygienists] and advanced manufacturing)

Springboard—a job that easily leads into a career path (such as human resource assistants,

computer support specialists, and bookkeepers)

Static—a job that is lower wage and does not lend itself to career growth (more common in the

manufacturing field but also includes some health-care jobs, such as physician’s assistant

and pharmacy tech; Lamback, Gerwin, & Restuccia, 2018).

The mismatch between industry needs and the supply of middle-skill workers is rooted in the

acceleration of technological change and globalization (Burrows et al., 2014; National Skills

Coalition, 2017). Employer survey reports indicate that some of hardest middle-skill positions to

fill are in careers with high lifetime values, such as health care, computer and mathematical

operations, and technical sales and management (Burrows et al., 2014). To compensate for this

unmet need, employers are “upskilling” their job postings (e.g., requiring a 4-year degree even

when unnecessary for a job) to find employees with the necessary technical (e.g., computer

literacy) and soft (e.g., work ethic) skills (Burrows et al., 2014).

The Labor Market Return to Middle-Skill Training In general, the middle-skill literature, as with the broader literature on the returns to education,

shows that more education is associated with higher earnings (Abel & Deitz, 2014; Belfield &

Bailey, 2017; Carnevale, Rose, & Cheah, 2011). In a 2011 study, Carnevale et al. found the

median lifetime earnings potential of a middle-skill employee to be $1.7 million, which is more

than half a million higher than their counterparts with only a high school diploma. A separate,

more recent study showed that the return on investment for associate degrees is 15% (Abel &

Deitz, 2014) or $4,640–$7,160 per year compared with entering college but not completing a

degree or certificate (Belfield & Bailey, 2017).

Literature that examines variation in these returns shows that students generally earn more

when completing longer duration credentials and credentials in fields related to science,

technology, engineering, and mathematics (Bosworth, 2010; Dadgar & Trimble, 2016; Jacobson,



LaLonde, & Sullivan, 2005; Prince & Jenkins, 2005).4,5 Some evidence also exists that certain

degree-granting institutions provide elevated returns relative to comparable credentials earned

at other institutions (e.g., for-profit compared with public or nonprofit institutions; Cellini &

Chaudhary, 2012; Deming, Yuchtman, Abulafi, Goldin, & Katz, 2016).

Slightly more than one

fourth of certificates

(27%6) are granted by

degree-granting

institutions. Recent and

sharp declines in state

funding for community

colleges have limited

public, degree-granting

institution enrollment

opportunities (e.g.,

greater restrictions on

the times that classes

are offered; McFarland

et al., 2019; Mettler,

2014). Although the parallel increases in available non-credit-bearing postsecondary

alternatives may help ease unmet market demand for middle-skill training, related research

may indicate that these alternatives do not serve students well and, more worrisome still,

provide substandard services to disproportionately underprivileged populations (Deming et al.,

2013; Mettler, 2014). For example, research shows that for-profit colleges expressly recruit

students from low-income backgrounds, charge tuition as much as five times higher than

students pay at community colleges, encourage students to take out large loans, and provide

equivalent or worse labor market outcomes than their public or nonprofit counterparts (Cellini

& Chaudhary, 2012; Darolia, Koedel, Martorell, Wilson, & Perez‐Arce, 2015; Deming et al., 2012,

2013; Kutz, 2010; Lang & Weinstein, 2013).

4 Bosworth (2010) highlighted differences in returns by the length of the certificate program and showed that “longer term certificates have significantly higher labor market value than short-term certificates because of their greater technical and academic rigor” (p. ii). Dadgar and Weiss (2012), using administrative data from Washington state, similarly found that associate degrees and longer term certificates have a positive impact on earnings. 5 An important part of the literature is whether there is an economic impact of participation in a higher education program without attaining a degree or certificate (Bailey, Kienzl, & Marcotte, 2004; Kane & Rouse, 1995), known as the “sheepskin” effect. We did not explore that distinction in this study. Results are varied (Belfield & Bailey, 2017). 6 See Table 320.10 at https://nces.ed.gov/programs/digest/d17/tables/dt17_320.10.asp.

https://nces.ed.gov/programs/digest/d17/tables/dt17_320.10.asp



In the present study, the

principal distinguishing

characteristic between the

programs of interest was the

ability to accrue transferable

and stackable credits from an

accredited, degree-granting

program. In the data we used,

non-credit-bearing programs

commonly led to certificates

for nurse’s aide, computer

operations, cardiopulmonary

resuscitation, business skills,

and cosmetology, along with licenses for medical assistant, cosmetology, insurance, emergency

medical technician, real estate, truck driving, and registered nurse.7 Whereas credit-bearing

programs may yield some of the same credentials, they are, by definition, associated with the

accrual of academic course credits and frequently culminate with an associate degree. Herein,

our central focus is on directly testing differences in the returns to terminal, non-credit-bearing

programs (whether they derived from public, nonprofit, or for-profit programs) compared with

more traditional, credit-bearing programs.

Methods

To estimate the relationship between career program type and the labor market trajectories of

participants following completion, we used the NLSY97 and an individual fixed-effects

regression analysis estimation strategy (based on Cellini & Chaudhary, 2012) to estimate the

association of program completion with changes in two labor market outcomes: employment,

measured using a binary variable that takes a value of 1 if employed and 0 otherwise, and

quarterly earnings calculated using average hourly wage × weekly hours worked.8 The specific

regression model that we used is in Equation 1.

7 At least two other paths to pursuing middle-skill jobs are worth mentioning but were outside the scope of this study. First, stackable-credential programs provide a hybrid between traditional non-credit-bearing certificate programs and credit-bearing academic programs by tying short-term credentials to more extended career pathways (Austin, Mellow, Rosin, & Seltzer, 2012). There are, however, at the moment, empirical challenges to distinguishing such programs, and current evidence shows that their usage is low (Belfield & Bailey, 2017). Second, apprenticeships also provide a path toward middle-skill jobs. In the past 5 years, the U.S. Department of Labor (DOL) has engaged in a significant transformation of national apprenticeship programs, expanding their reach in terms of the number of workers and beyond traditional industries and occupations (see https://www.dol.gov/featured/apprenticeship/grants for recent DOL grant programs). However, these are recent innovations and, during the time period examined in this study, apprenticeships were limited in scope (Klor de Alva & Schneider, 2018). 8 Appendix B also presents estimates for hours and wage outcome variables separately. These outcomes were combined in the estimates presented in the main text on earnings as earnings = wages × hours.

https://www.dol.gov/featured/apprenticeship/grants



𝑦𝑖𝑐𝑡 = α1+ α2𝑁𝑜𝐻𝑆𝑖𝑐𝑡 + α3𝐻𝑆𝑖𝑐𝑡 + α4𝐶𝑟𝑒𝑑𝑖𝑡𝑖𝑐𝑡 + α5𝐵𝐴𝑖𝑐𝑡 + α6𝑋𝑖𝑐𝑡 + γ𝑖 + δ𝑡 + θ𝑐𝑡 + ε𝑖𝑐𝑡 (1)

In this model,

𝑦𝑖𝑐𝑡 represents the outcome of interest, employment,9 or the natural log earnings for each

individual i in county c and year t;

𝑁𝑜𝐻𝑆𝑖𝑐𝑡 takes on a value of 1 for any individual i and county c whose highest level of

educational attainment in year t is less than a high school diploma or a General Educational

Development (GED) certificate;

𝐻𝑆𝑖𝑐𝑡 takes on a value of 1 for any individual i and county c whose highest level of educational

attainment in year t is a high school diploma or GED;

𝐶𝑟𝑒𝑑𝑖𝑡𝑖𝑐𝑡 takes on a value of 1 for any individual i in county c whose highest level of

educational attainment in year t includes a credit-bearing postsecondary certificate or an

associate degree (or any other credit-bearing credential that is designed to be completed in

less than 4 years and prior to attaining a bachelor’s degree); and

𝐵𝐴𝑖𝑐𝑡 takes on a value of 1 for any individual i in county c whose highest level of educational

attainment in year t includes a bachelor of arts or science postsecondary degree.

The omitted category is a non-credit-bearing certificate or licensure program.

The primary innovation of this report relied on the individual fixed effects (yi), which were used

to identify changes in employment and earnings that accrued to NLSY97 respondents after

completing credit-bearing programs compared with non-credit-bearing programs while holding

constant persistent differences between individuals, such as ability or motivation. Therefore, if

we are concerned that students select into non-credit-bearing programs instead of credit-

bearing programs because they are less able or motivated, if that ability or motivation factor is

consistent in an individual’s life, then these effects will effectively control for that.

The additional controls integrated in the model are for situations in which program selection

may change across time. Specifically, to account for variation in program availability and

context that changes across time, we integrated year fixed effects (δ𝑡) and county by year

trends (θ𝑐𝑡). To account for individuals’ different likelihoods to choose training programs,

whether they were single or married, as well as whether they have children, we controlled for a

host of individual by year factors, such as marital status, the number of children, and age by

race and age by gender fixed effects (𝑥𝑖𝑐𝑡).

9 Because employment is a 0/1 dummy variable, the employment estimates are all linear probability models.



One of the most critical identification challenges associated with using an individual fixed

effects strategy, such as Equation 1, is associated with any systematic, unobserved, and time-

varying differences in the employment or earnings trajectories between individuals who pursue

credit- and non-credit-bearing course taking. For example, individuals with an ongoing strong

connection to the labor market may prefer coursework that will be more targeted and less time

consuming, as typical of non-credit-bearing pathways. These individuals may also exhibit

smaller changes in earnings from before to after credential attainment because of their

preexisting workforce connections. In contrast, individuals who experience a disruption in their

labor market connection (e.g., through reduced hours or lower wages) may be more likely to

pursue credit-bearing options. They may then exhibit a larger increase in earnings from before

to after credential attainment because of the drop in earnings just prior to attainment.

To address these concerns, we integrate two controls into 𝑥𝑖𝑐𝑡 for a respondents’ participation

in work in any period during which the individual may have been concurrently enrolled in

school. These controls include one parameter for concurrent enrollment in a non-credit-bearing

program and a second for enrollment in a credit-bearing program.10 These controls are

designed to absorb variation associated with systematic differences in the propensity to work

during credit- and non-credit-bearing training, related to the concept of Ashenfelter’s Dip

(1978) associated with systematic differences in pre-training work behavior. We also estimate

Equation 1 separately for hours and wages to examine the extent to which any differential

return to earnings is driven by hourly wage.

The error term, ε𝑖𝑐𝑡, represented the variation that remained unexplained after accounting for

the parameters in the model. We estimated these models using ordinary least squares

regression and Huber-White clustering of standard errors at the state of residence at age 18 to

account for systematic variation in outcomes or participation experience by program (Huber,

1967; White, 1980).

The primary coefficient of interest in this model was α4, which indicates the average

differences in the employment and earnings outcomes between individuals whose highest

educational attainment was a credit-bearing credential or an associate degree compared with a

non-credit-bearing certificate or license.

We also present exploratory analyses that examined the extent to which attaining a credit-

bearing credential or associate degree as one’s highest educational attainment by age 30,

10 If the respondent reports concurrent credit- and non-credit-bearing enrollment, we code those as credit-bearing enrollment because evidence suggests that a credit-bearing program is the more long-term and potentially more time-consuming commitment.



versus a non-credit-bearing credential or license, affected one’s likelihood to ever attain a

bachelor’s degree. The model we used for this strategy is in Equation 2.

𝑦𝑖𝑐 = β1+ β2𝐶𝑟𝑒𝑑𝑖𝑡𝑖𝑐 + β3𝑋𝑖𝑐 + ε𝑖𝑐 (2)

In this model, the principal coefficient of interest was β2, which indicates the extent to which

credit-bearing programs were associated with a greater propensity to attain a bachelor’s degree

by age 30. The control variables (𝑋𝑖𝑐) included gender and race/ethnicity, as well as marital

status and the number of children at first attainment (or age 18 if no high school graduation or

higher attainment recorded). We also included fixed effects for age and county of residence at

first attainment (or age 18 if no high school graduation or higher attainment recorded).

Data

The data that we used for our analysis was the NLSY97, which provided longitudinal,

participant-level information on about 9,000 nationally representative respondents starting in

1997 (when respondents were ages 12–18) and continuing through 2016 (when respondents

were ages 31–37). Across this period, there were 17 interview rounds, and all data were self-

reported. We restricted the sample to 6,093 respondents, which was balanced to span ages

18–30 for each individual. We further restricted the sample to only those who did not earn a

bachelor’s degree within 6 years of high school graduation or GED attainment.

The NLSY97 contains two related strands of questioning associated with schooling and training

question stems that we used to create the educational attainment variables central to the

analysis. In Figure 1, we share the logic used to construct our educational attainment variables,

including attainment of a credit-bearing credential or associate degree and a non-credit-bearing

credential or license. This language was adapted from the question stems that start the

respective schooling and training sections in the NLSY97. The Bureau of Labor Statistics changed

the schooling and training questions stems slightly across the 17 rounds, but each time, they

followed a consistent logic. Appendix A provides additional detail on the manner in which we

used questions on participant labor market participation to construct our employment and

earnings dependent variables.



Figure 1. The Flow of Schooling and Training Questions in the NLSY97 Used to Identify

Educational Attainment Type and Date of Completion

Table 1 presents descriptive details on the gender, race/ethnicity, marital status, the number of

children, parental education, parental wealth, employment, and earnings of the NLSY97

respondents in our analysis sample. These statistics show that the sample was approximately

half female and more than 60% non-White, and few respondents (19%) at age 18 had a parent

with 4 or more years of college. The family makeup of respondents at the time of highest

educational attainment indicated that about 16% were married, and about 31% had a child.

About half of the respondents (47%) had no postsecondary training by age 30. Also, at age 30,

about 82% of the respondents were employed, and the average earnings (adjusted to 2016

dollars) were $37,888. As a point of comparison, these earnings were about 319% of the federal

poverty line for a single-family household in 2016. Taken together, these characteristics

indicate that the sample was disproportionately representative of underprivileged respondents

in the overall NLSY97 because of our restriction of the sample to include only those individuals

who did not graduate with a bachelor’s degree within 6 years of high school graduation.

If certificate, license, or certificate of completion,

Respondent earned a credit-bearing credential or degree as of reported receipt date (earliest date if more than one credential earned).

What diploma, degree, or certificate have you received from [school name]?

If yes,

Have you been enrolled in regular school? What type of school were you enrolled in? (Regular school is one that offers an academic diploma or degree. Not included as regular school are training at a technical institution, license trade programs, and so on, unless the credits obtained are transferrable to a regular school and could count toward an academic diploma or degree.)

Respondent earned a non-credit-bearing credential as of the date the respondent stopped attending the training.

Did you receive a certificate, license, or degree from this program? What type?

If answered,

If two-year associate degree, vocational or technical certificate or higher

What other types of schooling and training have you had, excluding the regular schooling we have already talked about? Some sources of occupational training programs include business schools, cosmetology schools, nursing courses, apprenticeship programs, vocational or technical institutes or schools, correspondence courses, company or military training, employer training programs, and night schools. We are interested in training programs you took to make it easier to find a job, improve your job skills, or learn a new job.



Table 1. NLSY97 Sample Characteristics

N Weighted

mean

percentage

Gender: Female 2,917 49%

Gender: Male 3,176 51%

Race/ethnicity: American Indian, Eskimo, or Aleut 31 1%

Race/ethnicity: Asian or Pacific Islander 64 2%

Race/ethnicity: Black or African American 1,864 15%

Race/ethnicity: Hispanic 1,463 13%

Race/ethnicity: White 2,593 67%

Race/ethnicity: Other 78 2%

Family characteristics: Marital status at time of highest educational attainment 793 16%

Family characteristics: At least one child at time of highest educational attainment 2,129 31%

Family characteristics: At least one parent with at least 4 years of college 912 19%

Highest educational attainment: No high school diploma/GED 555 8%

Highest educational attainment: High school diploma/GED 2,423 39%

Highest educational attainment: Non-credit-bearing certificate or license 1,992 33%

Highest educational attainment: Credit-bearing certificate or associate degree 676 12%

Highest educational attainment: Bachelor’s or higher degree 447 8%

Outcomes: Employment at age 30 4,884 82%

Outcomes: Earnings at age 30 5,761 $37,888

Note. The total sample included 6,093 individuals, which was restricted from the larger sample of 8,984 NLSY97

respondents to include individuals who did not attain a bachelor’s degree within 6 years of their high school

graduation. Earnings represent 2016 dollars. See Appendix A for detail on the weighting procedure.

Descriptive Changes in the Outcomes of Interest: Employment and Earnings

Our analysis examined two outcomes related to educational attainment: employment and

earnings. In this section, we look descriptively at the trajectories of these outcomes in two ways.

First, we plot the outcomes for people at different ages grouped by the highest level of education

they ever attained by age 30. Second, we focus on the years just before and after attainment.

These depictions suggest that credit-bearing programs are more likely to enhance an individual’s

labor market outcomes (particularly earnings) relative to non-credit-bearing programs.

Employment

The likelihood of employment at some point in a year was approximately equal among

respondents who completed credit-bearing and non-credit-bearing programs until their late



20s. At age 24, the likelihood of employment was 88% for respondents whose highest

educational attainment was a credit-bearing credential as well as for respondents whose

highest educational attainment was a non-credit-bearing credential. At age 30, the likelihood of

employment diverged just slightly between these groups to 88% and 84%, respectively (see

Figure 2).

Figure 2. Percentage Employed by Age and Highest Educational Attainment

Note. The total sample included 6,093 NLSY97 respondents who did not attain a bachelor’s degree within 6 years

of high school graduation or obtaining a GED. We applied custom weights and based profiles on respondents’

highest educational attainment. See Appendix A for detail on the weighting procedure.

Other changes that occur during an individual’s life-span may explain these descriptive findings

related to differences in the types of people who choose each path. Consider child rearing as an

example of such a change: Individuals are more likely to enroll in programs that they perceive

to be more challenging when they do not have children (or have fewer children); also, the

presence of children may reduce an individual’s likelihood of employment unrelated to their

educational attainment, especially among mothers (Gough & Noonan, 2013; Moore & Waite,

1977; Sibulkin & Butler, 2005). Therefore, these descriptive changes may be an artifact of

omitted variables that are simultaneously changing throughout a person’s life-span and

codetermined with educational attainment but are not a by-product of attainment.

To focus on the change in outcomes that we observed from the attainment of credentials, we

next present graphs that depict the years just prior to and following program completion (see

Figure 3). These depictions show the differences in employment likelihood that respondents

60%

65%

70%

75%

80%

85%

90%

95%

100%

18 19 20 21 22 23 24 25 26 27 28 29 30

Per

cen

tage

Em

plo

yed

Age

No high school diploma High school diploma

Non-credit-bearing credential Credit-bearing credential

Four-year or higher degree



experienced in the 3 years prior to credit-bearing and non-credit-bearing educational attainment

and the 4 years following attainment. In this figure, Year 0 represents the year of completion.11

The red vertical line presented in Year -1 depicts the final full year before program completion

and after which we would anticipate the divergence in employment likelihood.

The findings from this second descriptive investigation reinforced our interpretation of the

evidence presented in Figure 3. Although the employment likelihood was quite similar between

credit-bearing and non-credit-bearing program pathways prior to completion, in the years

immediately following completion, respondents who attained a credit-bearing certificate or a 2-

year degree were more likely to be employed.

Figure 3. Percentage Employed by Years Before and After Educational Attainment


of high school graduation or obtained a GED and were between the ages of 18 and 30 three years before and after

their highest educational attainment. We applied custom weights and based profiles on respondents’ highest

educational attainment. See Appendix A for detail on the weighting procedure.

Earnings

The descriptive earnings detail was consistent with and even larger in magnitude than the

descriptive employment detail. Specifically, Figure 4 shows that earnings were approximately

equivalent for respondents who would ultimately complete credit-bearing and non-credit-

11 Completion may have occurred at any point in the calendar year—as early as January or as late as December. Program participation may have occurred through all of Year -1 (e.g., through continuous enrollment in a 2-year program) or not at all (e.g., if participation in a short-term certificate program was wholly contained in Year 0).

60%

65%

70%

75%

80%

85%

90%

95%

100%

-3 -2 -1 0 1 2 3

Per

cen

tage

Em

plo

yed

Years Before and After Educational Attainment

Credit-bearing certificate/Less-than-four-year degree

Non-credit-bearing certificates/licenses




bearing programs throughout their mid-20s but diverged by age 30. For example, at age 24,

respondents whose highest attainment was from a credit-bearing certificate or 2-year degree

program earned an average of $34,619 per year, whereas respondents whose highest

attainment was from a non-credit-bearing certificate or licensure program earned an average

of $36,608 per year, a surplus of about $2,000 per year that accrued to respondents who

ultimately completed a non-credit-bearing program.12 However, by age 30, the groups earned

an average of $43,949 and $34,619, respectively—indicating that respondents who completed

credit-bearing programs attained about $10,000 more per year by age 30 (see Figure 4).

Figure 4. Earnings by Age and Highest Educational Attainment


of high school graduation or obtained a GED. We applied custom weights and based profiles on respondents’

highest educational attainment. See Appendix A for detail on the weighting procedure.

When we looked at earnings in the years before and after educational attainment, we saw data

consistent with the nature of employment: relatively flat earnings prior to attainment and a rise

in earnings after attainment (see Figure 5). The trend among respondents who attained non-

credit-bearing credentials showed an overall gain from Year -3 to Year +3, but it was flatter than

the curves for the other two categories.

12 Related to the way the figure was constructed, respondents may or may not have earned the noted credential (certificate or degree) by age 24. Respondents were grouped by the highest level of education ever attained, which may have been attained at a later age.

$-

$10,000

$20,000

$30,000

$40,000

$50,000

$60,000

18 19 20 21 22 23 24 25 26 27 28 29 30

Year

ly E

arn

ings

Age

No high school diploma High school diploma

Non-credit-bearing credential Credit-bearing credential




Figure 5. Earnings by Year Before and After Educational Attainment


of high school graduation or obtained a GED and were between the ages of 18 and 30 three years before and after

their highest educational attainment. We applied custom weights and based profiles on respondents’ highest

educational attainment. See Appendix A for detail on the weighting procedure.

Impact Findings

To examine the differential labor market returns to credit-bearing and non-credit-bearing

attainment, we used an individual fixed-effects strategy to identify changes in employment and

earnings that accrued to NLSY97 respondents upon completion. Our estimates showed that

credit-bearing programs did not yield a significant increase in the likelihood of employment

relative to non-credit-bearing programs (Table 2, findings column 1). Our estimates did,

however, show that credit-bearing programs increased yearly earnings by about 17% relative to

the $33,000 average earnings of respondents who completed non-credit-bearing programs and

who were employed, representing a change of about $5,500 (Table 2, findings column 2).13

13 Beyond the focus on the relative return associated with credit- and non-credit-bearing attainment, these results also showed significant improvements in the employment and earnings that respondents receive after completing a bachelor’s degree or higher degree-based program.

$10,000

$20,000

$30,000

$40,000

$50,000

$60,000

-3 -2 -1 0 1 2 3

Earn

ings

Years Before and After Educational Attainment

Credit-bearing certificate/Less-than-four-year degree

Non-credit-bearing certificates/licenses




Table 2. Regression Estimates of Employment Likelihood and Natural Log Earnings by

Educational Attainment

(1)

Employed

(2)

Natural log yearly earnings

No high school diploma or GED -0.082***

(0.014)

-0.313***

(0.046)

High school diploma or GED attainment -0.026***

(0.008)

-0.135***

(0.027)

Non-credit-bearing certificates/licenses (reference category)

Credit-bearing certificate/associate degree 0.017

(0.012)

0.153***

(0.044)

4-year or higher degree 0.083***

(0.017)

0.411***

(0.064)

Marital status -0.012**

(0.006)

0.028

(0.018)

Number of children -0.037***

(0.005)

-0.104***

(0.009)

Enrolled in School -0.037***

(0.006)

-0.259***

(0.013)

Enrolled in Training 0.011**

(0.006)

-0.035*

(0.020)

Constant 0.754***

(0.043)

8.480***

(0.319)

Observations 79,134 59,313

Number of individuals 6,093 6,015

R-squared 0.070 0.264

Note. Standard deviations are in parentheses. The total sample included 6,093 NLSY97 respondents who did not attain a bachelor’s degree within 6 years of high school graduation or obtained a GED. The models included the following fixed effects: individual, year, gender*age, race*age, and year trend*county. We applied custom weights and clustered standard errors at the state of residence at age 18. See Appendix A for detail on the weighting procedure. *p < .10. **p < .05. ***p < .01.

Hours and Wages

To examine the extent to which changes in earnings are associated with fundamental

differences in the relationship to work that individuals in credit- and non-credit-bearing

credentialing programs have, we present an analysis of our estimates separately on hours and

wages. These analyses indicated that although the earnings differences were largely driven by

changes in the number of hours worked, significant differences occurred in the wages that

respondents who completed credit-bearing training earned relative to respondents who

completed non-credit-bearing programs. Specifically, respondents who completed credit-



bearing programs worked about 100 more hours each year and earned about 9.2% more (Table

3).

Table 3. Regression Estimates of Yearly Hours Worked and Natural Log Hourly Wage by


(1)

Yearly hours

(2)

Natural log hourly wage


(37.24)

-0.056**

(0.024)


(26.49)

-0.061***

(0.013)


Credit-bearing certificate/associate degree 99.93**

(46.89)

0.088***

(0.031)


(64.42)

0.132***

(0.031)

Marital status -7.294

(18.47)

0.053***

(0.008)


(10.33)

-0.009

(0.007)


(15.77)

-0.072***

(0.009)

Enrolled in Training -17.17

(17.97)

-0.019*

(0.009)

Constant 774.3***

(226.7)

2.205***

(0.117)

Observations 76,163 59,313

Number of individuals 6,079 6,015

R-squared 0.168 0.200

Note. Standard errors in parentheses. The total sample included 6,093 NLSY97 respondents who did not attain a

bachelor’s degree within 6 years of high school graduation or obtained a GED. The models included the following fixed

effects: individual, year, gender*age, race*age, and year trend*county. We applied custom weights and clustered

standard errors at the state of residence at age 18. See Appendix A for detail on the weighting procedure.

*p < .10. **p < .05. ***p < .01.



Subgroup Analyses

Prior research indicated that individuals from low-wealth families, non-White individuals, and

individuals who were the first in their families to attend college were significantly less likely to

go directly from high school into a 4-year college program compared with their more

economically advantaged peers (McFarland et al., 2019). We next examined the stability of

these estimates across these more vulnerable subgroups.

The findings among our subgroups of interest show that employment results were consistent

across subgroups (Table 4). The results on earnings, on the other hand, indicate that respondents

from low-wealth families (in the lowest tercile of the wealth distribution when the respondent

was age 18) and non-White respondents did not experience significant earnings increases for

completing a credit-bearing program compared with completing a non-credit-bearing program

(Table 5, findings columns 2 and 3). We also found that the changes in hours and wages for low-

wealth and non-White respondents associated with credit-bearing credentials were not

statistically significant, as shown in Appendix B.

Table 4. Regression Estimates of Employment Likelihood Among Vulnerable Subgroups

(1)

Overall

(2)

Low Wealth

(3)

Non-White

(4)

Parent without

bachelor’s degree


(0.014)

-0.098**

(0.037)

-0.073

(0.045)

-0.079***

(0.015)


(0.008)

-0.056***

(0.020)

-0.019

(0.024)

-0.027***

(0.009)


Credit-bearing certificate/associate

degree

0.017

(0.012)

0.002

(0.042)

0.003

(0.031)

0.016

(0.015)


(0.017)

0.151**

(0.066)

0.082

(0.063)

0.066**

(0.027)

Marital status -0.012**

(0.006)

-0.014

(0.021)

0.013

(0.015)

-0.012

(0.008)


(0.005)

-0.037***

(0.012)

-0.035***

(0.010)

-0.037***

(0.005)


(0.006)

-0.031**

(0.012)

-0.032**

(0.012)

-0.035***

(0.006)

Enrolled in Training 0.011**

(0.006)

0.012

(0.018)

0.037**

(0.015)

0.012*

(0.006)

Constant 0.754***

(0.043)

1.080***

(0.147)

0.760***

(0.128)

0.751***

(0.085)



Observations 79,134 20,089 45,444 60,836

Number of individuals 6,093 1,547 3,500 4,685

R-squared 0.070 0.145 0.141 0.081

Note. Standard deviations are in parentheses. The total sample included 6,093 NLSY97 respondents who did not attain a




*p < .10. **p < .05. ***p < .01.

Table 5. Regression Estimates of Natural Log Yearly Earnings Among Vulnerable Subgroups

(1)

Overall

(2)

Low Wealth

(3)

Non-White

(4)

Parent without

bachelor’s degree


(0.046)

-0.232**

(0.109)

-0.141

(0.157)

-0.314***

(0.048)


(0.027)

-0.122**

(0.057)

0.000

(0.087)

-0.149***

(0.029)


Credit-bearing certificate/associate

degree

0.153***

(0.044)

0.101

(0.092)

0.274

(0.184)

0.156***

(0.044)


(0.064)

0.355

(0.23)

0.338**

(0.151)

0.370***

(0.109)

Marital status 0.028

(0.018)

0.087

(0.052)

-0.012

(0.049)

0.030

(0.022)


(0.009)

-0.092***

(0.022)

-0.05

(0.041)

-0.104***

(0.012)


(0.013)

-0.210***

(0.047)

-0.254***

(0.047)

-0.210***

(0.021)

Enrolled in Training -0.035*

(0.020)

-0.066

(0.052)

-0.102**

(0.040)

-0.038*

(0.021)

Constant 8.48***

(0.319)

7.525***

(0.345)

8.235***

(0.521)

8.551***

(0.309)

Observations 59,313 14,306 33,219 45,566


R-squared 0.264 0.305 0.376 0.254

Note. Standard deviations are in parentheses. The total sample included 6,093 NLSY97 respondents who did not attain a






*p < .10. **p < .05. ***p < .01.

Exploratory Analyses

Students may select into different education pathways for a multitude of reasons. These could

include differences in the potential outcomes associated with programs’ labor market earnings.

However, students also may consider any additional flexibility that credit-bearing pathways

provide, such as students’ ability to build on the credits they earned in the pathway toward

eventual attainment of a bachelor’s degree at a later point.

To investigate differences in the utility of the credit- and non-credit-bearing pathways in terms

of their flexibility, we followed the main results just presented with additional, exploratory

analyses. These analyses investigated differences in the flexibility of educational attainment by

examining the likelihood that students who started credit and non-credit-bearing pathways

eventually earned a bachelor’s degree.

The short-term employment or earnings gains associated with a program are unlikely to

capture the full range of benefits that individuals consider when selecting their educational

pursuits. Other potential benefits that individuals likely consider include some potential

signaling value associated with a degree and long-term earnings, both of which may accrue

because of the flexibility that a program may provide, especially considering the rapid evolution

of present-day labor markets. One hallmark of flexibility is associated with the ability to transfer

or stack credentials, typical in credit-bearing programs, which provides the opportunity to build

from preassociate certificates to associate degrees and then to bachelor’s degrees.

To test the extent to which this specific type of flexibility was associated with stacking from an

initial credit-bearing certificate or associate degree to a bachelor’s degree, we estimated the

differential propensity to attain a bachelor’s degree by age 30 between respondents whose first

educational attainment was from a credit-bearing program compared with respondents whose

first educational attainment was from a non-credit-bearing program (see Equation 2 and the

results in Table 7). With these findings, we show that respondents whose initial postsecondary

educational attainment was from a credit-bearing program were 11 percentage points more

likely to attain a bachelor’s degree by age 30 than respondents whose initial postsecondary

educational attainment was from a non-credit-bearing program.14

14 We also estimated results for each vulnerable subgroups. These subgroup analyses showed that (similar to the subgroup analyses relevant to the main results) vulnerable populations experience similar propensities to leverage a credit-bearing degree to attain a bachelor’s degree. See Appendix C.



Table 6. Regression Estimates of First Attainment in Credit-Bearing Programs Compared With

Non-Credit-Bearing Programs on the Propensity to Attain a Bachelor’s Degree by Age 30

Propensity to attain a bachelor’s degree

Credit-bearing certificate/associate degree (first attainment) 0.111***

(0.014)

Male -0.021*

(0.011)

White -0.01

(0.014)

Marital status at first attainment (or age 18) 0.002

(0.015)

Number of children at first attainment (or age 18) -0.021***

(0.007)

Constant 0.02

(0.088)

Age at first attainment (or age 18) fixed effects Yes

County at first attainment (or age 18) fixed effects Yes

Observations 2,859

R-squared 0.224


of high school graduation or obtained a GED. We applied custom weights and clustered standard errors at the

state of residence at age 18. See Appendix A for detail on the weighting procedure.

*p < .10. **p < .05. ***p < .01.

Conclusion

In this report, we showed that credit-bearing programs have significant positive returns on

students’ earnings relative to the returns that accrue to individuals who complete non-credit-

bearing postsecondary training programs. The magnitude of the earnings difference that accrues

to students who complete credit-bearing programs is on the order of $5,500 per year. We found

that this increase in earnings is associated in both an increase in hours as well as an increase in

the hourly wage. In addition, our research presented results from exploratory analyses that

provide modest evidence that credit-bearing programs are indeed more portable (i.e., used to

support ultimate bachelor’s degree attainment).

There are a number of policy implications for these findings, but further research is needed to help

sort out the efficacy of those policies. First, the higher returns to credit-bearing credentials signal a

need for increased support for these pathways and greater understanding of why individuals

choose non-credit-bearing programs. If a barrier to credit-bearing pathways is financial cost,



additional means-tested education subsidies may be necessary to ensure that all students can

contribute to the future labor market in accordance with their ability. Second, approaching this

from a different angle, if the differential in returns is caused by the hypothesized terminal nature of

non-credit-bearing programs, this could signal a need to examine potential mechanisms for better

connecting careers and educational pathways for students. This type of approach is reflected in

current discussions on middle skills, including the discourse on stackable credentials (Austin et al.,

2012; Belfield & Bailey, 2017; Bohn & McConville, 2018), and calls to connect apprenticeship

training to college credits (Klor de Alva & Schneider, 2018; McCarthy, Palmer, & Prebil, 2017).

In interpreting the results from this study, there are relevant considerations associated with the

systematic influence of the field of study, types of credentials, and individual background on

respondent outcomes. To examine these interaction effects, employers, states, and workforce

regulation boards should consider gathering ongoing evidence on the experiences and

outcomes of diverse students’ in a range of training programs to enhance their understanding

of the ability for a range of programs to serve them well. These data should aim to include

information on the roles that credit-bearing and non-credit-bearing training programs play in

facilitating or impeding students’ abilities to meet their education and career goals and succeed

in life. Without careful data collection, evaluation, and support services, training programs and

their stakeholders may undermine the longer-term successes of the students that they aim to

provide.



References

Abel, J. R., & Deitz, R. (2014). Do the benefits of college still outweigh the costs? Current Issues

in Economics & Finance, 20(3), 1–12. Retrieved from

https://www.newyorkfed.org/medialibrary/media/research/current_issues/ci20-3.pdf

Ashenfelter, O. (1978). Estimating the effect of training programs on earnings. The Review of

Economics and Statistics, 60(1), 47–57.

Austin, J. T., Mellow, G. O., Rosin, M., & Seltzer, M. (2012). Portable, stackable credentials: A

new education model for industry-specific career pathways. Columbus, OH: McGraw-Hill

Research Foundation. Retrieved from https://jfforg-prod-

prime.s3.amazonaws.com/media/documents/Portable_Stackable_Credentials.pdf

Bailey, T., Kienzl, G., & Marcotte, D. E. (2004). The return to a sub-baccalaureate education: The

effects of schooling, credentials and program of study on economic outcomes.

Washington, DC: U.S. Department of Education, National Assessment of Vocational

Education. Retrieved from https://www2.ed.gov/rschstat/eval/sectech/nave/subbac-

ed.pdf

Belfield, C., & Bailey, T. (2017). The labor market returns to sub-baccalaureate college: A review

(CAPSEE Working Paper). New York, NY: Center for Analysis of Postsecondary Education

and Employment. Retrieved from

https://ccrc.tc.columbia.edu/media/k2/attachments/labor-market-returns-sub-

baccalaureate-college-review.pdf

Bohn, S., & McConville, S. (2018). Stackable credentials in career education at California

community colleges. Retrieved from

Bosworth, B. (2010). Certificates count: An analysis of sub-baccalaureate certificates.

Indianapolis, IN: Complete College America. Retrieved from

https://files.eric.ed.gov/fulltext/ED536837.pdf

Burrows, J., Young, A., Restuccia, D., Fuller, J., & Raman, M. (2014). Bridge the gap: Rebuilding

America’s middle skills. Cambridge, MA: Harvard Business School. Retrieved from

https://www.hbs.edu/competitiveness/Documents/bridge-the-gap.pdf

Carnevale, A. P., Rose, S. J., & Cheah, B. (2011). The college payoff: Education, occupations,

lifetime earnings. Washington DC: Georgetown University, Center on Education and the

Workforce. Retrieved from https://cew.georgetown.edu/cew-reports/the-college-

payoff/#full-report

https://www.newyorkfed.org/medialibrary/media/research/current_issues/ci20-3.pdf

https://jfforg-prod-prime.s3.amazonaws.com/media/documents/Portable_Stackable_Credentials.pdf

https://jfforg-prod-prime.s3.amazonaws.com/media/documents/Portable_Stackable_Credentials.pdf

https://www2.ed.gov/rschstat/eval/sectech/nave/subbac-ed.pdf

https://www2.ed.gov/rschstat/eval/sectech/nave/subbac-ed.pdf

https://ccrc.tc.columbia.edu/media/k2/attachments/labor-market-returns-sub-baccalaureate-college-review.pdf

https://ccrc.tc.columbia.edu/media/k2/attachments/labor-market-returns-sub-baccalaureate-college-review.pdf


https://www.hbs.edu/competitiveness/Documents/bridge-the-gap.pdf

https://cew.georgetown.edu/cew-reports/the-college-payoff/#full-report

https://cew.georgetown.edu/cew-reports/the-college-payoff/#full-report



Carnevale, A. P., Smith, N., & Strohl, J. (2013). Recovery: Job growth and education

requirements through 2020. Washington, DC: Georgetown University, Center on

Education and the Workforce. Retrieved from

https://repository.library.georgetown.edu/bitstream/handle/10822/559311/Recovery2

020.FR.Web.pdf?sequence=1

Carnevale, A. P., Strohl, J., Ridley, N., & Gulish, A. (2018). Three educational pathways to good

jobs: High school, middle skills, and bachelor’s degree. Washington, DC: Georgetown

University, Center on Education and the Workforce. Retrieved from

https://cew.georgetown.edu/cew-reports/3pathways/

Cellini, S. R., & Chaudhary, L. (2012). The labor market returns to a for-profit college education

(NBER Working Paper No. 18343). Cambridge, MA: National Bureau of Economic

Research. Retrieved from https://www.nber.org/papers/w18343.pdf

Dadgar M., & Trimble, M. J. (2016). Labor market returns to sub-baccalaureate credentials: How

much does a community college degree or certificate pay? Education Evaluation and

Policy Analysis, 37(4), 399–418.

Dadgar, M., & Weiss, M. J. (2012). Labor market returns to sub-baccalaureate credentials: How

much does a community college degree or certificate pay? (CCRC Working Paper No. 45).

New York, NY: Teachers College, Columbia University, Community College Research

Center. Retrieved from https://files.eric.ed.gov/fulltext/ED533520.pdf

Darolia, R., Koedel, C., Martorell, P., Wilson, K., & Perez‐Arce, F. (2015). Do employers prefer

workers who attend for‐profit colleges? Evidence from a field experiment. Journal of

Policy Analysis and Management, 34(4), 881–903.

Deming, D. J., Goldin, C., & Katz, L. F. (2012). The for-profit postsecondary school sector: Nimble

critters or agile predators? The Journal of Economic Perspectives, 26(1), 139–163.

Deming, D., Goldin, C., & Katz, L. F. (2013). For-profit colleges. The Future of Children, 23(1),

137–163.

Deming, D. J., Yuchtman, N., Abulafi, A., Goldin, C., & Katz, L. F. (2016). The value of

postsecondary credentials in the labor market: An experimental study. American

Economic Review, 106(3), 778–806.

Fuller, J., & Raman, M. (2017). Dismissed by degrees: How degree inflation is undermining U.S.

competitiveness and hurting America’s middle class. Cambridge, MA: Harvard Business

School; Washington, DC: Grads of Life; Dublin, Ireland: Accenture. Retrieved from

http://www.hbs.edu/managing-the-future-of-work/Documents/dismissed-by-degrees.pdf

https://repository.library.georgetown.edu/bitstream/handle/10822/559311/Recovery2020.FR.Web.pdf?sequence=1

https://repository.library.georgetown.edu/bitstream/handle/10822/559311/Recovery2020.FR.Web.pdf?sequence=1

https://cew.georgetown.edu/cew-reports/3pathways/

https://www.nber.org/papers/w18343.pdf


http://www.hbs.edu/managing-the-future-of-work/Documents/dismissed-by-degrees.pdf



Gough, M., & Noonan, M. (2013). A review of the motherhood wage penalty in the United

States. Sociology Compass, 7(4), 328–342.

Handel, M. F. (2017). What are middle skills? Issues in Science & Technology, 33(2), 6–8.

Huber, P. J. (1967). The behavior of maximum likelihood estimates under nonstandard

conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and

Probability, 1(1), 221–233.

Jacobson, L., LaLonde, R., & Sullivan, D. G. (2005). Estimating the returns to community college

schooling for displaced workers. Journal of Econometrics, 125(1–2), 271–304.

Kane, T. J., & Rouse, C. E. (1995). Labor-market returns to two-and four-year college. American

Economic Review, 85(3), 600–614.

Klor de Alva, J., & Schneider, M. (2018). Apprenticeships and community colleges: Do they have

a future together? Washington, DC: American Enterprise Institute. Retrieved from

https://www.aei.org/wp-content/uploads/2018/05/Apprenticeships-and-Community-

Colleges.pdf

Kutz, G. D. (2010). For-profit colleges: Undercover testing finds colleges encouraged fraud and

engaged in deceptive and questionable marketing practices (Testimony before the

Committee on Health, Education, Labor, and Pensions, U.S. Senate; GAO-10-948T).

Washington, DC: U.S. Government Accountability Office. Retrieved from

https://www.gao.gov/assets/130/125197.pdf

Lamback, S., Gerwin, C., & Restuccia, D. (2018). When is a job just a job—and when can it

launch a career? The real economic opportunities of middle-skill work. Boston, MA: JFF.

Retrieved from https://jfforg-prod-

prime.s3.amazonaws.com/media/documents/ResumeDataBook6.pdf

Lang, K., & Weinstein, R. (2013). The wage effects of not-for-profit and for-profit certifications:

Better data, somewhat different results. Labour Economics, 24, 230–243.

Le, V-N., Yang, M., & Simko, C. (2017). Findings from the National Education and Employment

Survey Wave 1 and wave 2. Chicago, IL: University of Chicago, NORC. Retrieved from

https://www.norc.org/PDFs/NEES/Findings%20from%20NEES%20Waves%201%20and%

202_%20April%202017%20FINAL.pdf

https://www.aei.org/wp-content/uploads/2018/05/Apprenticeships-and-Community-Colleges.pdf

https://www.aei.org/wp-content/uploads/2018/05/Apprenticeships-and-Community-Colleges.pdf

https://www.gao.gov/assets/130/125197.pdf

https://jfforg-prod-prime.s3.amazonaws.com/media/documents/ResumeDataBook6.pdf

https://jfforg-prod-prime.s3.amazonaws.com/media/documents/ResumeDataBook6.pdf

https://www.norc.org/PDFs/NEES/Findings%20from%20NEES%20Waves%201%20and%202_%20April%202017%20FINAL.pdf

https://www.norc.org/PDFs/NEES/Findings%20from%20NEES%20Waves%201%20and%202_%20April%202017%20FINAL.pdf



Marr, B. (2019, July 15). The future of work: 5 important ways jobs will change in the 4th

industrial revolution. Forbes.com. Retrieved from

https://www.forbes.com/sites/bernardmarr/2019/07/15/the-future-of-work-5-

important-ways-jobs-will-change-in-the-4th-industrial-revolution/#12efdb2754c7

McCarthy, M. A., Palmer, I., & Prebil, M. (2017). Connecting apprenticeship and higher

education: Eight recommendations. Washington, DC: New America. Retrieved from

https://www.newamerica.org/education-policy/policy-papers/eight-recommendations-

connecting-apprenticeship-and-higher-ed/

McFarland, J., Hussar, B., Zhang, J., Wang, X., Wang, K., Hein, S., . . . Barmer, A. (2019). The

condition of education 2019 (NCES 2019-144). Washington, DC: U.S. Department of

Education, Institute of Education Sciences, National Center for Education Statistics.

Retrieved from https://nces.ed.gov/pubs2019/2019144.pdf

Mettler, S. (2014, March 20). Equalizers no more: Politics thwart colleges’ role in upward

mobility. The Chronicle of Higher Education. Retrieved from

https://www.chronicle.com/article/equalizers-no-more/

Moore, K. A., & Waite, L. J. (1977). Early childbearing and educational attainment. Family

Planning Perspectives, 9(5), 220–225.

National Skills Coalition. (2017). United States’ forgotten middle. Washington, DC: Author.

Retrieved from https://www.nationalskillscoalition.org/resources/publications/2017-

middle-skills-fact-sheets/file/United-States-MiddleSkills.pdf

Prince, D., & Jenkins, D. (2005). Building pathways to success for low-skill adult students:

Lessons for community college policy and practice from a statewide longitudinal tracking

study. New York, NY: Teachers, College, Columbia University, Community College

Research Center. Retrieved from https://ccrc.tc.columbia.edu/publications/low-skill-

adults-policy.html

Sibulkin, A. E., & Butler, J. S. (2005). Differences in graduation rates between young black and

white college students: Effect of entry into parenthood and historically black

universities. Research in Higher Education, 46(3), 327–348.

White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test

for heteroskedasticity. Econometrica: Journal of the Econometric Society, 48(4),

817–838.

https://www.forbes.com/sites/bernardmarr/2019/07/15/the-future-of-work-5-important-ways-jobs-will-change-in-the-4th-industrial-revolution/#12efdb2754c7

https://www.forbes.com/sites/bernardmarr/2019/07/15/the-future-of-work-5-important-ways-jobs-will-change-in-the-4th-industrial-revolution/#12efdb2754c7

https://www.newamerica.org/education-policy/policy-papers/eight-recommendations-connecting-apprenticeship-and-higher-ed/

https://www.newamerica.org/education-policy/policy-papers/eight-recommendations-connecting-apprenticeship-and-higher-ed/

https://nces.ed.gov/pubs2019/2019144.pdf

https://www.chronicle.com/article/equalizers-no-more/

https://www.nationalskillscoalition.org/resources/publications/2017-middle-skills-fact-sheets/file/United-States-MiddleSkills.pdf

https://www.nationalskillscoalition.org/resources/publications/2017-middle-skills-fact-sheets/file/United-States-MiddleSkills.pdf

https://ccrc.tc.columbia.edu/publications/low-skill-adults-policy.html

https://ccrc.tc.columbia.edu/publications/low-skill-adults-policy.html



Appendix A. Description of the NLSY97 Data

Overview of the Data Source For this study, we used NLSY97 public-use and restricted-use geographic data. We obtained the

public-use data through the NLS Investigator, an online tool that allows users to select and

download specific variables. The restricted-use data used for this study included respondents’

state and county of residence. A total of 8,984 NLSY97 respondents participated in the first

interview when respondents were between ages 12 and 18. At the time of this study, the 17th

interview round from 2015–16 included the most recent data available. We constructed a

yearly dataset, with one record per respondent for each calendar year between ages 18 and 30.

Sample Characteristics

We constructed the following characteristics for each respondent: race/ethnicity, age, marital

status, number of children, parents’ education level and level of family wealth, and county of

residence. In this appendix, we discuss the details associated with constructing each variable.

Race/Ethnicity

Respondents reported their race, ethnicity, and date of birth during the first interview. The

NLSY97 race categories included American Indian, Eskimo, or Aleut; Asian or Pacific Islander;

Black or African American; White; and some other race. We categorized respondents who

reported being a Hispanic ethnicity as Hispanic, regardless of their reported race.

Age

Age was defined using respondents’ year of birth.

Marital Status and Number of Children

Respondents’ marital status and number of children varied across years in the dataset. We

obtained marital status in a given year using a combination of respondents’ reported year of

their first marriage, reported year that their first marriage ended, and reported legal marital

status at the time of each interview. The dates for more than one marriage were not available.

If respondents reported that their first marriage ended, we categorized these respondents as

married again if they reported that their marital status was married at the time of the interview

after the end of their first marriage. Respondents who reported being married again after their

first marriage were then categorized as not married if they reported being divorced or

widowed. We used the birth year of respondents’ biological and adopted children to calculate

the number of children that respondents had in a given year.



Parents’ Education Level and Level of Family Wealth

We created two characteristics based on respondents’ parents that we collected at the first

interview round. To determine parents’ education level, we used the highest grade level

completed by respondents’ residential mother and father. We identified a mother or father

with at least 4 years of college completed as having completed a bachelor’s degree.

We used respondents’ reported household net worth at the first interview if respondents were

not independent to determine whether respondents were from a low-wealth family. We then

adjusted the household net worth to 2016 dollars. We categorized respondents in households

in the bottom third of net worth as from a low-wealth family.

County of Residence

We collected respondents’ state and county of residence at each interview round. Therefore,

respondents’ state and county data were not available during a year when a respondent was

not interviewed. To obtain respondents’ state and county information for years when a

respondent was not interviewed or data was missing between ages 18 and 30, we assumed that

the respondent resided in the same state and county that was reported during the previous

interview.


Our principal independent variable of interest was educational attainment. To construct these

measures, we leveraged NLSY97 data on the years that respondents received each of the

following: (a) a high school diploma or GED, (b) a non-credit-bearing or credit-bearing

certificate or license, (c) an associate degree, or (d) a bachelor’s degree or higher. When

possible, we used NLSY97-created variables for the year that respondents reported receiving a

high school diploma, a GED, or a degree from regular schooling. These NLSY97-created variables

used information reported across all rounds for all respondents. If more than one degree of the

same type was reported, the earliest valid receipt date was presented.15

We used the NLSY97-created variables for the year that respondents obtained a high school

diploma or GED. No respondent had both a high school diploma receipt date and a GED receipt

date. We used the NLSY97-created date for when a respondent left high school if a respondent

was missing a diploma or GED receipt year (i.e., the respondent reported receipt of a high

school diploma or GED but did not provide a date of receipt).

15 See https://www.nlsinfo.org/content/cohorts/nlsy97/other-documentation/codebook-supplement/appendix-1-education-variable/date for more information about how NLSY97 created each date.

https://www.nlsinfo.org/content/cohorts/nlsy97/other-documentation/codebook-supplement/appendix-1-education-variable/date

https://www.nlsinfo.org/content/cohorts/nlsy97/other-documentation/codebook-supplement/appendix-1-education-variable/date



The year that respondents received their first non-credit-bearing certificate or license was

obtained using respondents’ report of attainment from a training program. The language in the

NLSY97 training questionnaire asked respondents to report on training rather regular

postsecondary schooling: “Other than high school, college, or university degree programs you

may have told me about earlier, since [date of last interview], have you attended any schooling,

courses, or training programs designed to help people find a job, improve their job skills, or

learn a new job? Some sources of occupational training programs include trade schools, nursing

courses, apprenticeships, vocational or technical institutes, company or military trainings,

correspondence courses, employer training programs, and other continuing education.”16 We

identified respondents who reported receiving a vocational certificate, a state license, a

certificate of completion, or other license from a training program as non-credit-bearing

certificate or license recipients. We defined the date of receipt as the date the respondent

stopped attending the training or last stopped attending the training if the respondent was

attending the training at the time of the interview because the date of receipt was not

collected. We retained the earliest date of receipt across all interview rounds. We did not use

the NLSY97-created variable for the date that respondents reported receipt of a training

certificate or license because the date presented reflects the most recent certificate or license,

not the earliest attainment.

Respondents reported receipt of an associate degree or a credit-bearing certificate in the

schooling questionnaire. Items in the schooling questionnaire asked respondents to report on

the following types of schools: 2-year colleges, community colleges, junior colleges, 4-year

colleges or universities, graduate schools, law schools, and medical schools. We obtained the

year that respondents received an associate degree from an NLSY97-created variable. An

NLSY97-created variable was not available for the receipt date of vocational or technical

certificates reported in the schooling questionnaire. Therefore, we examined respondents’

reported vocational or technical certificate receipt from regular schooling across all interview

rounds and retained the earliest date if more than one certificate was reported. We retained

the earliest associate degree or certificate receipt year.

We also used NLSY97-created variables to identify the year that respondents received a

bachelor’s or higher degree. The degree types that respondents reported on the schooling

questionnaire included bachelor’s degree, master’s degree, doctoral degree, and professional

degree. We retained the earliest receipt year across all degree types.

16 Taken from https://nlsinfo.org/content/cohorts/nlsy97/topical-guide/education/training; language changed slightly across interview rounds.

https://nlsinfo.org/content/cohorts/nlsy97/topical-guide/education/training



Employment and Earnings

Our principal outcomes of interest were employment and earnings. We obtained respondents’

employment status and total earnings in a given year using the information collected for each

job. At each interview, we collected employment and wage data for all jobs since respondents’

last interview. In addition, jobs had unique identification numbers, which allowed us to tracked

respondents’ jobs across interview rounds.

We used the start and stop years that respondents reported working at each job to determine

employment status during a given year. The year the respondent stopped working at the job is

the interview date if the respondent was still employed at the job at the time of the interview.

Therefore, we used the start and stop dates for a given job reported in the most recent

interview round. We identified a respondent as employed in a given year if the respondent was

employed with any job at any point between the start and stop years.

We used the NLSY97-created variables for reported hourly compensation and hours worked per

week for each job in each interview round to calculate the yearly earnings of respondents.

Respondents could report their compensation and hours worked for each job in various time

frequencies, including hourly, daily, weekly, monthly, and yearly. The NLSY97-created variables

transformed the compensation and hours worked reported at various frequencies to one

frequency that was comparable across all jobs. The compensation and hours worked used in

the NLSY97 calculation was the amount that respondents reported receiving at the time of the

employment stop date (i.e., the date the respondent was no longer employed by the employer

or the date of the interview if the respondent was still employed by the employer) if the job

lasted for more than 13 weeks. If the job lasted 13 weeks or less, the compensation and hours

worked reflect the amount that respondents reported at the start of the job. Compensation

included tips, bonuses, commissions, overtime, and incentive pay.

We made additional adjustments to the NLSY97-created variables for each job. If weekly

compensation or hours worked per week were missing for a given job and interview round, we

used the weekly compensation or hours calculated in the previous interview round or in the

subsequent round for the same job. Hours worked per week greater than 80 hours were top

coded to 80 hours per week. Hourly compensation amounts greater than $1,000 for a given job

were top coded to $999. Hourly compensation also was adjusted to 2016 dollars.

We converted the NLSY97-created hourly and weekly variables to obtain the total yearly

earnings used in this study. First, we multiplied hourly compensation and hours worked per

week to obtain earnings per week at each job. Then we calculated both earnings per month and

hours worked per month for a given job by multiplying the weekly amounts by the average

number of weeks in a month (4.33 weeks). We then summed respondents’ monthly earnings



for all months that a respondent was employed at each job within each year. The monthly

earnings used for a given month were for those months between the job start date or stop date

reported in the previous interview round if the job was ongoing and the stop date reported in

the given interview round. Finally, we summed respondents’ yearly earnings across all jobs.

Weights We downloaded the NLS custom weights for the specific set of respondents in our sample. The

weights allowed us to correct for oversampling, clustering, and nonresponse to the interview.

For more information about the NLS custom weighting program, see

https://www.nlsinfo.org/weights/custom-weighting-program-documentation.

https://www.nlsinfo.org/weights/custom-weighting-program-documentation



Appendix B. Findings on Wages and Hours

Table B1. Regression Estimates of Yearly Hours Worked Among Vulnerable Subgroups

(1)

Overall

(2)

Low

Wealth

(3)

Non-White

(4)

Parent without

bachelor’s degree


(37.24)

-263.3***

(83.20)

-293.9***

(104.80)

-299.8***

(39.53)


(26.49)

-168.8***

(58.99)

-102.2*

(51.77)

-128.2***

(25.33)


Credit-bearing certificate/associate degree 99.93**

(46.89)

-80.68

(101.30)

160.90

(95.94)

120.2**

(55.54)


(64.4)

385.6**

(177.0)

212.6*

(110.2)

359.1***

(84.2)

Marital status -7.29

(18.47)

68.70

(53.96)

-15.15

(57.92)

-12.19

(23.18)


(10.33)

-89.61***

(22.40)

-114.3***

(24.39)

-116.5***

(10.73)


(15.77)

-235.8***

(39.54)

-211.6***

(31.91)

-239.0***

(17.39)

Enrolled in Training -17.17

(17.97)

-43.84

(46.54)

-11.95

(63.38)

-11.98

(16.62)

Constant 774.3***

(226.7)

755.1**

(288.4)

655.9*

(334.3)

773.2***

(222.1)

Observations 76,163 19,438 43,938 58,613


R-squared 0.168 0.236 0.261 0.163





*p < .10. **p < .05. ***p < .01.



Table B2. Regression Estimates of Natural Log Hourly Wage Among Vulnerable Subgroups

(1)

Overall

(2)

Low

Wealth

(3)

Non-White

(4)

Parent without

bachelor’s degree

No high school diploma or GED -0.056**

(0.024)

-0.055

(0.046)

0.082

(0.075)

-0.082***

(0.028)


(0.013)

-0.054

(0.037)

0.012

(0.045)

-0.081***

(0.019)


Credit-bearing certificate/associate degree 0.088***

(0.031)

0.104

(0.062)

0.090

(0.117)

0.076**

(0.029)


(0.031)

0.192**

(0.085)

0.144

(0.088)

0.157***

(0.034)

Marital status 0.053***

(0.008)

0.071***

(0.021)

0.012

(0.030)

0.054***

(0.011)

Number of children -0.009

(0.007)

-0.015

(0.012)

0.022

(0.034)

-0.018*

(0.009)


(0.009)

-0.043**

(0.018)

-0.073*

(0.040)

-0.054***

(0.012)

Enrolled in Training -0.019*

(0.009)

-0.050**

(0.022)

-0.021

(0.035)

-0.021*

(0.011)

Constant 2.205***

(0.117)

1.903***

(0.126)

2.075***

(0.119)

2.079***

(0.172)

Observations 59,313 14,306 33,219 45,566


R-squared 0.200 0.275 0.264 0.206





*p < .10. **p < .05. ***p < .01.



Appendix C. Exploratory Subgroup Analysis Results

Table C1. Regression Estimates of First Attainment in Credit-Bearing Programs Compared

With Non-Credit-Bearing Programs on the Propensity to Attain a Bachelor’s Degree by Age 30

(1)

Overall

(2)

Low Wealth

(3)

Non-White

(4)

Parent without

bachelor’s

degree

Credit-bearing certificate/associate degree

(first attainment)

0.111***

(0.014)

0.081**

(0.032)

0.114***

(0.02)

0.109***

(0.015)

Male -0.021*

(0.011)

-0.051**

(0.025)

0

(0.015)

-0.024**

(0.011)

White -0.01

(0.014)

0.061

(0.042) NA

-0.003

(0.016)

Marital status at first attainment (or age 18) 0.002

(0.015)

0.009

(0.034)

0.003

(0.021)

0.022

(0.016)

Number of children at first attainment (or

age 18)

-0.021***

(0.007)

-0.026*

(0.014)

-0.018**

(0.008)

-0.021***

(0.007)

Constant 0.02

(0.088)

-0.018

(0.248)

-0.009

(0.258)

0.022

(0.099)

Age at first attainment (or age 18) fixed effects Yes Yes Yes Yes

County at first attainment (or age 18) fixed

effects

Yes Yes Yes Yes

Observations 2,859 656 1,582 2,196

R-squared 0.224 0.392 0.249 0.252


bachelor’s degree within 6 years of high school graduation or obtained a GED. We applied custom weights and

clustered standard errors at the state of residence at age 18. See Appendix A for detail on the weighting

procedure.

*p < .10. **p < .05. ***p < .01.

LOCATIONS

Domestic: Arlington, VA (HQ) | Sacramento and San Mateo, CA | Atlanta, GA | Chicago, IL | Indianapolis, IN | Waltham, MA

Frederick and Rockville, MD | Chapel Hill, NC | Cayce, SC | Austin, TX

International: Ethiopia | Haiti

12606_09/20

Established in 1946, with headquarters in

Arlington, VA, the American Institutes for

Research® (AIR®) is a nonpartisan, not-for-profit

organization that conducts behavioral and social

science research and delivers technical assis-

tance, both domestically and internationally, in

the areas of education, health, and the workforce.

For more information, visit www.air.org.

MAKING RESEARCH RELEVANT

AMERICAN INSTITUTES FOR RESEARCH

1400 Crystal Drive, 10th Floor

Arlington, VA 22202-3239 | 202.403.5000

www.air.org

http://www.air.org/